Merge pull request #338 from BiomedicalMachineLearning/newmana/support-python-versions

newmana · web-flow · commit 375bdb12c47b · 2025-10-20T15:51:45.000+10:00
Newmana/support python versions 3.11 and 3.12.
diff --git a/.github/workflows/python-package.yml b/.github/workflows/python-package.yml
@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [3.10.18]
+        python-version: [3.12.12]
 
     steps:
     - uses: actions/checkout@v2
diff --git a/.readthedocs.yml b/.readthedocs.yml
@@ -8,7 +8,7 @@ version: 2
 build:
   os: ubuntu-24.04
   tools:
-    python: "3.10"
+    python: "3.12"
 
 # Build documentation in the "docs/" directory with Sphinx
 sphinx:
diff --git a/CONTRIBUTING.rst b/CONTRIBUTING.rst
@@ -66,11 +66,18 @@ Ready to contribute? Here's how to set up `stlearn` for local development.
 
 3. Install your local copy into a virtualenv. This is how you set up your fork for local development::
 
-    $ conda create -n stlearn-dev python=3.10 --y
+    Run the following:
+    $ conda create -n stlearn-dev python=3.12 --y
     $ conda activate stlearn-dev
     $ cd stlearn/
     $ pip install -e .[dev,test]
 
+    If you get an error for louvain package on MacOS, make sure you have cmake installed first (if you have brew):
+    $ brew install cmake
+
+    You can also use conda to install these dependencies (after creating the environment):
+    $ conda install -c conda-forge louvain leidenalg python-igraph
+
    Or if you prefer pip/virtualenv::
 
     $ python -m venv stlearn-env
diff --git a/HISTORY.rst b/HISTORY.rst
@@ -2,6 +2,12 @@
 History
 =======
 
+1.2.0 (2025-10-20)
+------------------
+* Added support for Python 3.11 and 3.12.
+* Upgraded scanpy to 1.11 - clustering will be different.
+* Added more CCI tests.
+
 1.1.5 (2025-09-17)
 ------------------
 * Add Leiden clustering wrapper.
diff --git a/docs/installation.rst b/docs/installation.rst
@@ -13,7 +13,7 @@ Install by PyPi
 Prepare conda environment for stLearn
 ::
 
-    conda create -n stlearn python=3.10 --y
+    conda create -n stlearn python=3.12 --y
     conda activate stlearn
 
 **Step 2:**
diff --git a/docs/release_notes/1.2.0.rst b/docs/release_notes/1.2.0.rst
@@ -0,0 +1,7 @@
+1.1.5 `2025-10-20`
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. rubric:: Features
+* Added support for Python 3.11 and 3.12.
+* Upgraded scanpy to 1.11 - clustering will be different.
+* Added more CCI tests.
diff --git a/docs/release_notes/index.rst b/docs/release_notes/index.rst
@@ -1,6 +1,8 @@
 Release Notes
 ===================================================
 
+.. include:: 1.2.0.rst
+
 .. include:: 1.1.5.rst
 
 .. include:: 1.1.1.rst
diff --git a/pyproject.toml b/pyproject.toml
@@ -4,21 +4,23 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "stlearn"
-version = "1.1.5"
+version = "1.2.0"
 authors = [
     {name = "Genomics and Machine Learning lab", email = "andrew.newman@uq.edu.au"},
 ]
 description = "A downstream analysis toolkit for Spatial Transcriptomic data"
 readme = {file = "README.md", content-type = "text/markdown"}
 license = {text = "BSD license"}
-requires-python = "~=3.10.0"
+requires-python = ">=3.10,<3.13"
 keywords = ["stlearn"]
 classifiers = [
     "Development Status :: 5 - Production/Stable",
     "Intended Audience :: Developers",
     "License :: OSI Approved :: BSD License",
     "Natural Language :: English",
     "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
 ]
 dynamic = ["dependencies"]
 
@@ -34,6 +36,7 @@ dev = [
     "furo==2024.8.6",
     "myst-parser>=0.18",
     "nbsphinx>=0.9.0",
+    "types-tensorflow>=2.8.0",
     "sphinx-autodoc-typehints>=1.24.0",
     "sphinx-autosummary-accessors>=2023.4.0",
 ]
diff --git a/requirements.txt b/requirements.txt
@@ -1,15 +1,15 @@
-bokeh==3.7.3
-click==8.2.1
-leidenalg==0.10.2
-louvain==0.8.2
-numba==0.58.1
-numpy==1.26.4
-pillow==11.3.0
-scanpy==1.10.4
-scikit-image==0.22.0
-tensorflow==2.14.1
-keras==2.14.0
-types-tensorflow>=2.8.0
-imageio==2.37.0
-scipy==1.11.4
-scikit-learn==1.7.0
+bokeh>=3.7.0,<4.0
+click>=8.2.0,<9.0
+leidenalg>=0.10.0,<0.11
+louvain>=0.8.2
+numba>=0.58.1
+numpy>=1.26.0,<2.0
+pillow>=11.0.0,<12.0
+scanpy>=1.11.0,<2.0
+scikit-image>=0.22.0,<0.23
+tensorflow>=2.14.1
+keras>=2.14.0
+pandas>=2.3.0
+imageio>=2.37.0,<3.0
+scipy>=1.11.0,<2.0
+scikit-learn>=1.7.0,<2.0
diff --git a/stlearn/embedding/pca.py b/stlearn/embedding/pca.py
@@ -9,7 +9,7 @@ def run_pca(
     data: AnnData | np.ndarray | spmatrix,
     n_comps: int = 50,
     zero_center: bool | None = True,
-    svd_solver: str = "auto",
+    svd_solver: str = "arpack",
     random_state: int | RandomState | None = 0,
     return_info: bool = False,
     use_highly_variable: bool | None = None,
@@ -38,11 +38,11 @@ def run_pca(
         Passing `None` decides automatically based on sparseness of the data.
     svd_solver
         SVD solver to use:
-        `'arpack'`
+        `'arpack'` (the default - deterministic)
           for the ARPACK wrapper in SciPy (:func:`~scipy.sparse.linalg.svds`)
         `'randomized'`
           for the randomized algorithm due to Halko (2009).
-        `'auto'` (the default)
+        `'auto'`
           chooses automatically depending on the size of the problem.
     random_state
         Change to use different initial states for the optimization.
diff --git a/stlearn/preprocessing/normalize.py b/stlearn/preprocessing/normalize.py
@@ -1,6 +1,3 @@
-from collections.abc import Iterable
-from typing import Literal
-
 import numpy as np
 import scanpy
 from anndata import AnnData
@@ -12,8 +9,7 @@ def normalize_total(
     exclude_highly_expressed: bool = False,
     max_fraction: float = 0.05,
     key_added: str | None = None,
-    layers: Literal["all"] | Iterable[str] | None = None,
-    layer_norm: str | None = None,
+    layer: str | None = None,
     inplace: bool = True,
 ) -> dict[str, np.ndarray] | None:
     """\
@@ -48,18 +44,6 @@ def normalize_total(
     key_added
         Name of the field in `adata.obs` where the normalization factor is
         stored.
-    layers
-        List of layers to normalize. Set to `'all'` to normalize all layers.
-    layer_norm
-        Specifies how to normalize layers:
-        * If `None`, after normalization, for each layer in *layers* each cell
-          has a total count equal to the median of the *counts_per_cell* before
-          normalization of the layer.
-        * If `'after'`, for each layer in *layers* each cell has
-          a total count equal to `target_sum`.
-        * If `'X'`, for each layer in *layers* each cell has a total count
-          equal to the median of total counts for observations (cells) of
-          `adata.X` before normalization.
     inplace
         Whether to update `adata` or return dictionary with normalized copies of
         `adata.X` and `adata.layers`.
@@ -76,8 +60,7 @@ def normalize_total(
         exclude_highly_expressed=exclude_highly_expressed,
         max_fraction=max_fraction,
         key_added=key_added,
-        layers=layers,
-        layer_norm=layer_norm,
+        layer=layer,
         inplace=inplace,
     )
 
diff --git a/stlearn/wrapper/read.py b/stlearn/wrapper/read.py
@@ -11,8 +11,10 @@
 import pandas as pd
 import scanpy
 from anndata import AnnData
+from h5py import File
 from matplotlib.image import imread
 from PIL import Image
+from scanpy import read_csv
 
 import stlearn
 from stlearn.types import _BACKGROUND, _QUALITY
@@ -87,8 +89,6 @@ def Read10X(
 
     adata.uns["spatial"] = dict()
 
-    from h5py import File
-
     with File(path / count_file, mode="r") as f:
         attrs = dict(f.attrs)
 
@@ -369,7 +369,6 @@ def ReadMERFISH(
     coordinates = pd.read_excel(spatial_file, index_col=0)
     if coordinates.min().min() < 0:
         coordinates = coordinates + np.abs(coordinates.min().min()) + 100
-    from scanpy import read_csv
 
     counts = read_csv(count_matrix_file).transpose()
 
diff --git a/tests/test_CCI.py b/tests/test_CCI.py
@@ -12,6 +12,12 @@
 import stlearn.tl.cci.het_helpers as het_hs
 from tests.utils import read_test_data
 
+# Per line - which cells to annotate
+CELL_TYPE_ANNOTATIONS = ["CT1", "CT2", "CT3", "CT2", "CT1", "CT3", "CT2"]
+
+# 3 cell types: A,E -> CT1, B,G -> CT2, C,F -> CT3
+CELL_TYPE_LABELS = np.array(["CT1", "CT2", "CT3"])
+
 global adata
 adata = read_test_data()
 
@@ -212,15 +218,17 @@ def test_get_interactions(self):
             3 neighbours express receptor:
                 * One is cell type 1, two are cell type 2.
         """
-        cell_annots = [1, 2, 3, 2, 1, 3, 2]
-        cell_data = np.zeros((len(cell_annots), 3), dtype=np.float64)
-        for i, annot in enumerate(cell_annots):
-            cell_data[i, annot - 1] = 1
-        all_set = np.array([str(i) for i in range(1, 4)])
-        sig_bool = np.array([True] + ([False] * (len(cell_annots) - 1)))
-        l_bool = sig_bool
-        r_bool = np.array([False] * len(cell_annots))
-        r_bool[[3, 4, 6]] = True
+
+        # Create 0 matrix using the above annotations to create position.
+        # i.e. CT1 = 0.
+        cell_data = TestCCI.create_cci(CELL_TYPE_ANNOTATIONS, CELL_TYPE_LABELS)
+
+        # Create middle ligand interacting with 3 neighbour receptors.
+        sig_bool = np.array([True] + ([False] * (len(CELL_TYPE_ANNOTATIONS) - 1)))
+        ligand_boolean = sig_bool.copy()
+
+        receptor_boolean = np.array([False] * len(CELL_TYPE_ANNOTATIONS))
+        receptor_boolean[[3, 4, 6]] = True
 
         # NOTE that format of output is an edge list for each celltype-celltype
         # interaction, where edge list represents interactions between:
@@ -235,10 +243,10 @@ def test_get_interactions(self):
             cell_data,
             self.neighbourhood_bcs,
             self.neighbourhood_indices,
-            all_set,
+            CELL_TYPE_LABELS,
             sig_bool,
-            l_bool,
-            r_bool,
+            ligand_boolean,
+            receptor_boolean,
             0,
         )
 
@@ -253,6 +261,48 @@ def test_get_interactions(self):
             self.assertEqual(len(observed_edgesi), len(expect_edgesi))
             self.assertTrue(np.all(match_bool))
 
+    def test_get_interaction_matrix(self):
+        """Test getting the interaction matrix for cell type pairs."""
+
+        # Create 0 matrix using the above annotations to create position.
+        # i.e. CT1 = 0.
+        cell_data = TestCCI.create_cci(CELL_TYPE_ANNOTATIONS, CELL_TYPE_LABELS)
+
+        # Middle spot (A) is significant and expresses ligand
+        sig_bool = np.array([True] + ([False] * 6))
+        ligand_bool = sig_bool.copy()
+
+        # Neighbors D, E, G express receptor
+        receptor_bool = np.array([False] * len(CELL_TYPE_ANNOTATIONS))
+        receptor_bool[[3, 4, 6]] = True
+
+        # Get interaction matrix
+        int_matrix = het.get_interaction_matrix(
+            cell_data,
+            self.neighbourhood_bcs,
+            self.neighbourhood_indices,
+            CELL_TYPE_LABELS,
+            sig_bool,
+            ligand_bool,
+            receptor_bool,
+            cell_prop_cutoff=0.2
+        )
+
+        # Expected: CT1 (A) -> CT2 (D,G): 2 interactions, CT1 -> CT1 (E): 1 interaction
+        # Matrix is [CT1->CT1, CT1->CT2, CT1->CT3, CT2->CT1, ...]
+        self.assertEqual(int_matrix.shape, (3, 3))
+        self.assertEqual(int_matrix[0, 0], 1)  # CT1 -> CT1 (A->E)
+        self.assertEqual(int_matrix[0, 1], 2)  # CT1 -> CT2 (A->D, A->G)
+        self.assertEqual(int_matrix[0, 2], 0)  # CT1 -> CT3 None
+
+    @staticmethod
+    def create_cci(cell_annotations: list[str], unique_cell_type_labels):
+        cell_data = np.zeros((len(cell_annotations), len(unique_cell_type_labels)),
+                             dtype=np.float64)
+        for i, annot in enumerate(cell_annotations):
+            ct_index = np.where(unique_cell_type_labels == annot)[0][0]
+            cell_data[i, ct_index] = 1
+        return cell_data
+
     # TODO next things to test:
-    #   1. Getting the interaction matrix.
-    #   2. Getting the LR scores.
+    #   1. Getting the LR scores.
